Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
1. A multisensory change detection system comprising: at least one processor; and memory including instructions that, when executed by the at least one processor, cause the at least one processor to: determine a first distribution for a first set of signal data from a first combination of sensors in a plurality of sensors; determine a second distribution for a second set of signal data from a second combination of sensors in the plurality of sensors; estimate a divergence between the first distribution and the second distribution based at least in part on a count of the first combination of sensors, a count of the second combination of sensors, a count of the plurality of sensors, signal value differences between a plurality of signal values in the second set of signal data and a first plurality of nearest neighbor signal values in the first set of signal data, and signal value differences between a plurality of signal value in the second set of signal data and a second plurality of nearest neighbor signal values in the second set of signal data; determine whether the divergence exceeds a threshold; and generate an alert in response to the divergence exceeding the threshold.
2. The system of claim 1 , the memory further including instructions to: identify a subset combination of sensors from the plurality of sensors; receive a third set of signal data from the subset combination of sensors; and determine the threshold using a model trained with the third set of signal data.
This invention relates to a sensor-based system for monitoring and analyzing signals to determine operational thresholds. The system addresses the challenge of accurately identifying and adapting to varying environmental or operational conditions by dynamically adjusting thresholds based on sensor data. The system includes a plurality of sensors configured to generate signal data and a memory storing instructions for processing this data. The memory includes instructions to identify a subset of sensors from the plurality, receive signal data from this subset, and determine an operational threshold using a model trained with the subset's signal data. The model is trained to adapt to specific conditions or anomalies detected by the subset, improving accuracy and reliability. The system may also include a processor to execute the instructions and a communication interface to transmit the determined threshold to other components. This approach allows for real-time adjustments, reducing false positives and enhancing system performance in dynamic environments. The invention is particularly useful in applications requiring precise monitoring, such as industrial equipment, environmental sensing, or healthcare diagnostics.
3. The system of claim 2 , the memory further including instructions to: identify a second subset combination of sensors from the plurality of sensors; receive a fourth set of signal data from the second subset combination of sensors; normalize the third set of signal data and the fourth set of signal data as normalized set of signal data; and train the model using the normalized set of signal data.
4. The system of claim 3 , wherein the third set of signal data and the fourth set of signal data are from a selected time period.
The invention relates to a signal processing system designed to analyze and compare multiple sets of signal data over specific time periods. The system is particularly useful in applications where temporal alignment and comparison of signal data are critical, such as in telecommunications, sensor networks, or industrial monitoring. The system processes at least four sets of signal data, where the third and fourth sets are derived from a selected time period. This allows for precise temporal correlation between the signals, enabling accurate analysis of changes or patterns within that timeframe. The system may also include mechanisms to filter, amplify, or otherwise preprocess the signal data before comparison, ensuring high-quality input for subsequent analysis. Additionally, the system may incorporate a user interface or automated control to select the time period for the third and fourth sets of signal data, providing flexibility in analyzing different intervals. The system can then output the results of the comparison, which may include visualizations, statistical metrics, or alerts based on predefined thresholds. By focusing on a selected time period for the third and fourth sets of signal data, the system enhances the accuracy and relevance of the analysis, making it particularly valuable in scenarios where temporal dynamics are important. This approach ensures that the comparison is not affected by irrelevant or outdated data, improving decision-making and system performance.
5. The system of claim 3 , wherein the subset combination of sensors and the second subset combination of sensors share at least one sensor.
6. The system of claim 3 , wherein the model is a sliding-window model.
7. The system of claim 1 , wherein the divergence is estimated using signal value differences between a plurality of signal values in the first set of signal data and both a third plurality of nearest neighbor signal values in the first set of signal data and a fourth plurality of nearest neighbor signal values in the second set of signal data.
8. The system of claim 1 , the memory further including instructions to determine the threshold using a target false alarm rate, wherein the target false alarm rate is determined recursively using a ratio of false alarm rates to correct alarm rates.
9. The system of claim 1 , the memory further including instructions to: receive data from a third combination of sensors at a first period and a second period, the data representing sensor states; analyze the first and second periods to determine a difference between the sensor states of the third combination of sensors at the first period and the second period; and estimate the divergence using the determined difference.
10. The system of claim 9 , wherein the sensor states include at least one of a sensor location, a sensor angle, a sensor orientation, a sensor movement, or a sensor operational status.
11. A method for multisensory change detection system comprising: determining a first distribution for a first set of signal data from a first combination of sensors in a plurality of sensors; determining a second distribution for a second set of signal data from a second combination of sensors in the plurality of sensors; estimating a divergence between the first distribution and the second distribution based at least in part on a count of the first combination of sensors, a count of the second combination of sensors, a count of the plurality of sensors, signal value differences between a plurality of signal values in the second set of signal data and a first plurality of nearest neighbor signal values in the first set of signal data, and signal value differences between a plurality of signal value in the second set of signal data and a second plurality of nearest neighbor signal values in the second set of signal data; determining whether the divergence exceeds a threshold; and generating an alert in response to the divergence exceeding the threshold.
12. The method of claim 11 , further comprising: identifying a subset combination of sensors from the plurality of sensors; receiving a third set of signal data from the subset combination of sensors; and determining the threshold using a model trained with the third set of signal data.
This invention relates to sensor-based systems for monitoring and analyzing signals, particularly in applications where sensor data is used to detect anomalies or trigger actions. The problem addressed involves improving the accuracy and reliability of threshold-based decision-making in sensor networks by dynamically selecting and training models on specific subsets of sensor data. The method involves a system with multiple sensors generating signal data. A subset of these sensors is identified based on their relevance to a particular monitoring task. Signal data from this subset is then collected and used to train a machine learning model. The trained model determines an optimal threshold value, which is applied to subsequent sensor data to detect events or anomalies. This approach allows the system to adapt to changing conditions by focusing on the most informative sensors and refining the threshold dynamically. The invention improves upon prior systems by reducing noise and irrelevant data, leading to more accurate and context-aware threshold determinations. This is particularly useful in applications like industrial monitoring, environmental sensing, or healthcare, where sensor data variability can impact decision-making. The method ensures that the threshold is tailored to the specific subset of sensors being monitored, enhancing system performance.
13. The method of claim 12 , further comprising: identifying a second subset combination of sensors from the plurality of sensors; receiving a fourth set of signal data from the second subset combination of sensors; normalizing the third set of signal data and the fourth set of signal data as normalized set of signal data; and training the model using the normalized set of signal data.
14. The method of claim 11 , wherein the divergence is estimated using signal value differences between a plurality of signal values in the first set of signal data and both a third plurality of nearest neighbor signal values in the first set of signal data and a fourth plurality of nearest neighbor signal values in the second set of signal data.
15. The method of claim 11 , further comprising determining the threshold using a target false alarm rate, wherein the target false alarm rate is determined recursively using a ratio of false alarm rates to correct alarm rates.
16. The method of claim 11 , further comprising: receiving data from a third combination of sensors at a first period and a second period, the data representing sensor states; analyzing the first and second periods to determine a difference between the sensor states of the third combination of sensors at the first period and the second period; and estimating the divergence using the determined difference.
17. The method of claim 16 , wherein the sensor states include at least one of a sensor location, a sensor angle, a sensor orientation, a sensor movement, or a sensor operational status.
18. At least one non-transitory computer readable medium including instructions for a multisensory change detection system that when executed by at least one processor, cause the at least one processor to perform operations to: determine a first distribution for a first set of signal data from a first combination of sensors in a plurality of sensors; determine a second distribution for a second set of signal data from a second combination of sensors in the plurality of sensors; estimate a divergence between the first distribution and the second distribution based at least in part on a count of the first combination of sensors, a count of the second combination of sensors, a count of the plurality of sensors, signal value differences between a plurality of signal values in the second set of signal data and a first plurality of nearest neighbor signal values in the first set of signal data, and signal value differences between a plurality of signal value in the second set of signal data and a second plurality of nearest neighbor signal values in the second set of signal data; determine whether the divergence exceeds a threshold; and generate an alert in response to the divergence exceeding the threshold.
19. The at least one non-transitory computer readable medium of claim 18 , the operations further to: identify a subset combination of sensors from the plurality of sensors; receive a third set of signal data from the subset combination of sensors; and determine the threshold using a model trained with the third set of signal data.
This invention relates to sensor-based systems for detecting anomalies or events, particularly in industrial or environmental monitoring applications. The problem addressed is the need for accurate and adaptive threshold determination in sensor networks to improve detection reliability while reducing false positives. The system involves a plurality of sensors generating signal data, which is processed to detect events or anomalies. A key aspect is dynamically adjusting detection thresholds based on learned patterns from sensor data. The invention includes a non-transitory computer-readable medium storing instructions that, when executed, perform operations to identify a subset of sensors from the plurality, receive signal data from this subset, and determine an optimal threshold using a model trained with this subset data. The model is trained to adapt thresholds based on the specific characteristics of the selected sensor subset, improving detection accuracy for varying conditions or sensor configurations. The system may also include operations to preprocess sensor data, apply filtering techniques, and compare processed signals against the determined threshold to identify events. The model training process involves analyzing historical or real-time data from the sensor subset to establish thresholds that minimize false detections while maintaining sensitivity. This adaptive approach allows the system to function effectively across different sensor deployments and environmental conditions.
20. The at least one non-transitory computer readable medium of claim 19 , the operations further to: identify a second subset combination of sensors from the plurality of sensors; receive a fourth set of signal data from the second subset combination of sensors; normalize the third set of signal data and the fourth set of signal data as normalized set of signal data; and train the model using the normalized set of signal data.
21. The at least one non-transitory computer readable medium of claim 18 , wherein the divergence is estimated using signal value differences between a plurality of signal values in the first set of signal data and both a third plurality of nearest neighbor signal values in the first set of signal data and a fourth plurality of nearest neighbor signal values in the second set of signal data.
22. The at least one non-transitory computer readable medium of claim 18 , the operations further to determine the threshold using a target false alarm rate, wherein the target false alarm rate is determined recursively using a ratio of false alarm rates to correct alarm rates.
23. The at least one non-transitory computer readable medium of claim 18 , the operations further to: receive data from a third combination of sensors at a first period and a second period, the data representing sensor states; analyze the first and second periods to determine a difference between the sensor states of the third combination of sensors at the first period and the second period; and estimate the divergence using the determined difference.
24. The at least one non-transitory computer readable medium of claim 22 , wherein the sensor states include at least one of a sensor location, a sensor angle, a sensor orientation, a sensor movement, or a sensor operational status.
Unknown
March 30, 2021
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.